Instead of re-opening Bug 4419, here is a new entry.

Conf::Parser::pack_eval_method currently uses the following regexp:
  s/^\s*(?:['"](.*?)['"]|([\d\.:A-Za-z]+?))\s*(?:,\s*|$)//

1. The above allows most domain names to be given as arguments to eval rule
in unquoted form, except for domains with dashes. I propose we also allow
a minus in the character set, so that one can do:
  full T9 eval:check_dkim_valid(gmail.com, cc.yahoo-inc.com, yahoo.com)
instead of:
  full T9 eval:check_dkim_valid(gmail.com, 'cc.yahoo-inc.com', yahoo.com)

2. The current regexp allows argument with mismatched quotes like
'foo.com", and does now make it possible to specify a ' or " within
the quoted argument, e.g. "foo',bar" or 'foo",bar'.

Here is a proposed change to deal with both:

<<<
    while ($args =~ s/^\s*(?:['"](.*?)['"]|([\d\.:A-Za-z]+?))\s*(?:,\s*|$)//) {
      if (defined $1) {
        push @args, $1;
      }
      else {
        push @args, $2;
      }
    }

>>>
    local($1,$2,$3);
    while ($args =~ s/^\s* (?: (['"]) (.*?) \1 | ( [\d\.:A-Za-z-]+? ) )
                       \s* (?: , \s* | $ )//x) {
      if (defined $2) {
        push @args, $2;
      }
      else {
        push @args, $3;
      }
    }